Whither PDP-11/70 Updates?
You know, I'm glad you asked. Well, not glad, but also not unhappy that you asked. Just kind of a calm, lukewarm tapwater of feelings because you asked about this. Actually, no one has asked about this. I don't think anyone reads this blog, that's fine, I mostly do this because every now and then the weasels in my skull get loud enough that I need to vent the pressure, in blog form. And also by mixing my metaphors.
Where was I? Soon. I'll have another PDP-11/70 entry soon. I have plans. Such plans. I don't want to spoil it for you but I more or less wrapped the project up months ago and man am I bad at keeping up with blog posts. Besides, the true hotness these days is video, so what I clearly need to do is buy a GoPro, strap it to my forehead and just start soldering stuff on camera while rambling about technology and unboxing things and just kind of fetishizing computer objects and saying things like "can you believe that the processor Steve Jobs invented for your cellphone is 1000x more powerful than this here refrigerator sized computer" and all that jazz. It's about metrics, it's about eyeballs, it's about pretending I know what I'm talking about, it's all about the dopamine rush of Being Liked on the Internet. Give it to me.
OK, but what about the Motorola MDP-1000?
Quiet, you.
So what are you going to write about, then?
Well as it turns out one of the reasons I didn't write much about either of the above two things is that I got distracted. By stuff. Other projects. Turns out I'm easily distracted and in this grim meathook hellscape of a decade distractions are what keeps me from going out of my everloving mind. So I thought I'd do a brief recap of the distractions from 2021, and the distractions I have planned for 2022 and beyond. I kind of hope to write a bit about them over the next year, perhaps even with some kind of regularity. I owe it to you, the one person who reads this blog. Thank you for your patronage, it means a lot to us.
So without further ado, here's a rundown of interesting new acquisitions in 2021, in roughly chronological order, and some of the projects that may or may not have resulted from them.
Three Rivers PERQ 2 Workstation
You may or may not be familiar with the Three Rivers PERQ workstation, but suffice it to say it is one of the more interesting workstations to come out of the early 80s. Designed in 1979, released to the world in 1981 and more or less forgotten today, the PERQ was a very ambitious machine: 768x1024 bitmapped graphics, with a microprogrammed bitslice processor, a megabyte of memory, 24mb local storage, ethernet, and a mouse-like pointing deivce. All in a package that'd fit (mostly) under your desk and for a very reasonable price. It was well ahead of its time in 1981 but unfortunately for a variety of reasons it failed to find a foothold in the market.
This is one of the rarest PERQ systems that 3RCC actually shipped. The PERQ 2 was an attempted refinement of the PERQ 1 workstation c. late 1982 with the intent of reducing the amount of noise the system generated. The PERQ 1 was fairly loud, housing a 14" hard drive and a pile of fans with very little in the way of acoustic design. Having one running under your desk made for a loud, hot workspace. The PERQ 2 attempted to solve this by using a smaller 8" drive and enclosing everything in a ton of metal with fewer fans. Unfortunately this had the side effect of making it difficult to service and much more prone to overheating. And it was still pretty loud.
Not many of this system were made. It was pretty quickly superseded by the PERQ T1 and T2 systems designed by ICL in the UK. This particular unit was in use at Carnegie-Mellon University, apparently as a 3mbit to 10mbit Ethernet bridge, running Accent. (The 3mbit hardware is, alas, no longer in the system.) I'd never even seen a picture of one of these before I acquired the specimen pictured above. It's in excellent shape (the monitor has a couple of dings that'll buff right out) and I hope to restore it to operating condition later this year. The biggest unknown is the 8" Micropolis hard drive. It's an uncommon drive with a nonstandard interface that's apparently prone to failure. Good times. Wouldn't be the first time a Micropolis drive has let me down when I needed it most.
ICL PERQ 3A Workstations
The PERQ 3A. Well, two of 'em. |
The PERQ 3A workstation was the last gasp of the PERQ line, c. 1985 or so, and never made it to market. Rather than use a custom bitslice processor it uses a conventional 16Mhz Motorola 68020, hoping to compete with Sun et. al in the then hot UNIX workstation market. The systems pictured to the left are prototypes, and I'm hoping to get at least one of them to work. I was able to recover most of the data from the hard drives in them and it looks like one of them may have source code for the 3B's operating system -- PNX, a UNIX System V R2 port done by ICL.
Just a few ECOs here... |
DEC RF08/RS08 Fixed-Head Disc Drive
The RF08/RS08 system is an early storage device for PDP-8 systems. A single RF08 controller could control up to four RS08 fixed-head disc drives. Each RS08 drive contained 128 fixed heads, addressing 256KW (12-bit word size) of storage on a single fixed platter in the drive. Thus a complete RF08 system could provide 1 megaword of storage, which was nothing to sneeze at in 1967 on a system as small as the PDP-8. The disk assemblies are extremely fragile and very difficult to run due to an interesting design decision: when the drive is spun down, the heads rest on the platter. This makes it all too easy for the heads to damage the surface of the drive, especially after sitting at rest for 50 years.
The rack pictured above shows the first powerup of the controller and drive logic after restoring the power supply. My intent is to emulate the RS08 side of things with a modern microcontroller and use the original RF08 controller with it. I hope to make this my primary project starting early this year, it would be really cool to get it running with my PDP-8/I system. My ultimate goal is to use it to run TSS/8 on my PDP-8/I. The RF08 was used as primary storage and swap space for TSS/8 systems, due to its fast response time (no waiting for heads to seek!). If I can get an emulated drive working, I may then attempt to restore an actual disk unit, but it's going to be a lift and I'm not really expecting success with it. Just too many fiddly parts. The two disk assemblies I have have had all their cabling cut, none of the platters are in particularly great shape and I have no idea what else is in store for me should I attempt to take it on.
RS08 Platter (note the cut ribbon cables) |
AMI Jukebox
OK, this isn't a computer but I always wanted a jukebox and this showed up locally for a really ridiculously reasonable price and I had to have it. It plays 45s, it's neat, it's complete, it needs a complete overhaul but it was made in my hometown and I love it and it is mine.
Wicat 150WS Workstation
In mid July of 2021 a kind fellow out in Port Angeles offered me a huge lot of old workstation hardware -- Early Sun, SGI and Wicat systems with tons of spares and piles of documentation. I drove out there in a rented cargo van with a friend and we loaded the van to the gills and returned with some real gems. Wicat was one of the first companies to build a system around the Motorola 68000, and the 150WS is one of their earliest systems, c. 1982 or so. At introduction it ran UNIX as well as Wicat's own MCS operating system. As time went on, Wicat focused more on MCS and was moderately successful into the early 90s.
I got three incomplete 150WS systems and a box of spare parts and was able to make one reliably working system out of them. The major issue is the multibus backplanes -- these had been sitting in a garage for a couple of decades and picked up a fair amount of corrosion that I was unable to completely eliminate in two of the systems. It should be possible (if painful) to replace the P1/P2 connectors in the slots with new (or new-old-stock) parts but I haven't yet attempted this.
I'm still working on cataloguing all the hardware and software that came with the systems, I've scanned a few thousand pages of documentation and I still have a few piles left to go.
SGI IRIS 2400 Turbo Workstations
As part of the Port Angeles haul, I also acquired two SGI IRIS 2400 Turbo systems and enough spares to populate a third system. This is a system I've been looking for for a long, long time. Early SGI hardware is extremely uncommon and to be offered this stuff was a dream come true.
The 2400 Turbo was introduced in 1985 and is based around the Motorola 68020. (The earlier non-turbo system was 68010 based, as were its predecessors, the 1000-series). It featured a geometry engine consisting of a pipeline of 12 custom VLSI processors dedicated to transforming, clipping, and scaling of 3D points, as well as a microcoded rasterization engine based around four AMD 2903 chips. The display supported up to 32 bitplanes, used for storing pixel and z-buffer data.
Both IRIS systems I received were in fairly decent condition but hadn't been run in years. One of the two was completely empty apart from backplane and power supply. In December I undertook an effort to get one of the two up and running, and thanks to the help of a friend I got the system passing diagnostics and booting GL2-W3.6 just about a week ago. I intend to write a more detailed post about this system later this year if all goes well, and I'd like to try to get the other system running as well.
Xerox 1100 ("Dolphin") and 1132 ("Dorado") Workstations
Xerox 1100 (Dolphin) on the right |
Xerox 1132 (Dorado) on the left |
Example Dorado board. Note the edge fingers on both sides! |
The Dorado's Base Board (front-end processor). Note the 6502 processor, how cute. |
The Dolphin and Dorado are two workstations that mere mortals shouldn't be allowed to look at, let alone possess. Both of these were intended as personal workstations for working in the Interlisp, Smalltalk, or Mesa programming languages and both are mostly forgotten -- overshadowed by the Alto and the Star workstations. Technically they were sold commercially as Interlisp and Smalltalk workstations -- brochures exist and they showed up in industry rags -- but most were used internally at Xerox PARC. Only a handful of each were made.
The 1132, aka the "Dorado" or "D1" was the first successor to the Alto, and design work began on it in 1975 in the CSL (Computer Science Laboratory) at Xerox PARC. It was intended to deal with some of the Alto's shortcomings: limited memory, relatively slow speed and lack of virtual memory. The first prototype ("Model 0") was complete and running in 1978, and the "Model 1", which improved on the original design based on lessons learned while developing the prototype, was running in spring of 1979. By 1983 there were around 75 Dorados in use at PARC as personal workstations for the programming staff. An interesting writeup of the history and design of the Dorado can be found here, it's a fun read and you should check it out.The Dorado was also available commercially as an Interlisp-D or Smalltalk-80 system, though I have no idea how many were sold -- it cannot have been many, and I suspect most went to universities and research firms.
The Dorado made up for the Alto's shortcomings, and then some. It was based around 10K ECL technology with a microcode clock speed of 16Mhz and provided for up to 8mb of memory. Execution was pipelined with an Instruction Fetch Unit (IFU) and under (extremely) optimal conditions could execute 1 Mesa PrincOp per microcode clock, or 16 million instructions per second. It was blisteringly fast for the time, and absolutely insane for a computer intended as a single-user, personal workstation. If you wanted to run Smalltalk quickly, this thing was what you needed.
Relative speed of Smalltalk-80, c. 1982, from here. |
The Dorado did have a few disadvantages: the use of ECL made it extremely power hungry, which meant that it also required a lot of cooling, which in turn made it too loud (and hot) to have in your office. Unlike the Alto, which lived under each engineer's desk at PARC, the Dorados had to be housed in air-conditioned computer rooms, with only the consoles located in programmers' offices.
The Dorado is more than likely the system that Steve Jobs and Bill Gates saw Smalltalk demonstrated on, not an Alto.
Cover of 1100 SIP Brochure |
Like the Alto and the Dorado it's based around a custom bitsliced, microcoded, task-oriented processor, and like other machines in the "D" series it's designed to execute Mesa bytecodes efficiently (though they were popular hosts for Interlisp-D and Smalltalk-80 as well). Like the Alto before it (and the Star after) it used standard TTL components. It contains a 24" Shugart SA4000 disk for local storage, up to 1.5mb of memory, bitmapped display, and both 3mbit and 10mbit Ethernet interfaces for talking to the world. (It is, in fact, the first system to use a 10mbit Ethernet interface, which had just become standardized at the time the 1100 was introduced.)
The 1100 is about the size of the Alto -- it'll fit under your desk and doesn't make too much noise. It was faster with more memory, and was more capable than the Alto (but a bit less so than the later Star.)
The 1100 and 1132 systems I acquired this year were both used at the Language Research and Development Center at the University of Pittsburgh. At this time I have no idea what exactly they were being used for there, but I'm hoping that some digging around may turn up more information. I have a acquired a console that can be used by either of these systems, but it's not yet come home.
At the moment I'm working on getting the 1100 to boot over the 3mbit Ethernet, using Ken Shirriff's 3mbit BeagleBone adapter for the hardware, and my IFS software to provide the services needed to do so. I have it successfully pulling down microcode from the network, but diagnostics run by the microcode indicate a failure on the memory controller that I'm working on resolving. I'm hoping to have things successfully booting by the time the console arrives, so I can use the network to back up the contents of the hard drive.
DEC TC01 DECtape Controller
I've been looking for a DECtape controller to use with my PDP-8/I for quite some time (as part of my "Get a PDP-8/I Running TSS/8" scheme) and one came up locally in trade for a different DECtape controller. This thing has everything -- blinkenlights, Diode-Transistor Logic, and lots and lots of flipchips. Not a single integrated circuit in the thing and it was a real joy to debug, even when it was confusing the hell out of me. After a couple of weeks of debugging this back in November it's up and running, paired with a very nice TU55 drive I traded a rather dirty TU56 for. I booted OS/8 from it late last year, the first time my PDP-8/I has run a real operating system in a very long time.
Now I need to find a Data Multiplexer (DM01) so that I can run both the TC01 and the RF08 at the same time. Then I will be living the dream.
I may also have a functional DW8E soon...
Conclusion
In conclusion, computers are a land of contrasts. I will be taking no questions at this time.
:) I'll read it.
ReplyDeleteI know all too well that it takes a lot of work to write these sorts of things. Looking forward to your next entries!
ReplyDeleteDefinitely been reading this blog. I am very excited about the PERQ acquisition. I first laid eyes on the PERQ at CMU and it was the first "3M" workstation I had ever seen. It is hard to describe the rush of awe seeing that for the first time. Of course, I used many impressive workstations after that (including Lisp Machines), but nothing had the same impact as that first encounter.
ReplyDelete